Covariance and Gain-based Federated Unscented Kalman Filter for Acoustic-Visual-Inertial Underwater Navigation

Modern mobile robots require precise and robust localization and navigation systems to achieve mission tasks correctly. In particular, in the underwater environment, where Global Positioning Systems (GPSs) cannot be exploited, the development of localization and navigation strategies becomes more challenging. The most exploited approaches are based on the use of the Doppler Velocity Log (DVL) sensor, which provides highly precise linear velocity estimates. Furthermore, payload sensors like Forward-Looking SONARs (FLSs) or optical cameras are employed for inspection and they can be used as effective support (or substitute) for the DVL in underwater navigation. When the DVL, the FLS and the camera are operative, considering that multiple speed measurements are available, sensor fusion strategies can increase the estimation performance. In particular, two Federated Unscented Kalman Filters (Federated UKFs) strategies are presented here, where the approaches, which are both parallel filtering strategies, differ on two aspects, namely the algorithm to compute the optimal state estimate and covariance matrix. In particular, Covariance-based and a Gain-based Federated UKFs will be evaluated and compared. With regard to the investigation of the here presented strategies, two autonomous missions were performed in Vulcano Island, Messina (Italy) with FeelHippo AUV, and the collected data have been employed for offline validation.